Davis Vaughan

I can see the value in both, and I kind of changed my mind after a bit of thinking. I added a branch that implements `rename()` with support for renaming...

If it seems like it works nicely, I'll use it. If not, I'll revert back to returning a `tibble`. I can include a warning, but the way it currently works...

One thing i worry about is being too noisy, especially when programming with this. I do the select() thing all the time inside package functions, knowing that it will drop...

Great question. The simple answer is, you can't! At least not yet. I've been thinking about this for awhile, and have termed them "continuous" vs "disjoint" filters. Continuous range filters...

It might be worth revisiting this if `vec_unchop()` has gotten faster https://github.com/tidyverse/dplyr/blob/dc3e0e6ccc54b049ecd89ca83455cdb101e6583d/R/join-rows.R#L54 Hmm, maybe not faster in some odd cases. This was from that example ``` r library(vctrs) index_flatten expression...

I think most of it comes from the casting and the assignment. I feel like we could probably make the assignment faster for at least the atomic cases, but that's...

This is not entirely true, see https://github.com/DavisVaughan/methodtest for a counterpoint where `...` is in the generic and we still get a warning

Side bar: Should `pack()` allow renaming in the tidyselection? That would avoid the `rename()` calls above. Not sure how that would interact with `names_sep` though - I guess `names_sep` would...

This is pretty nice: ``` r treatments_dt id variable treatment date #> 1: 1 1 A 1 #> 2: 3 1 A 2 #> 3: 2 2 B 3 #>...

I guess this is the closest we can get with `pivot_longer()` directly. Requiring names that match `names_sep` or `names_pattern` seems somewhat restrictive ``` r treatments %>% rename(A_treatment = A, B_treatment...